NRG-beamFoam: a simplified numerical FSI solver developed in OpenFOAM for modelling axial-flow in nuclear reactors

Abstract

Nuclear power is increasingly seen as a potential solution to the decarbonization of the energy sector in the coming decades. However, one of the main causes of downtime for current-generation nuclear reactors is the phenomenon of Grid-To-Rod-Fretting (GTRF) inside the reactor core. This is predominantly caused by Fluid-Structure Interaction (FSI), where the turbulent cooling water flow causes unwanted vibrational behaviour of the long and slender rods where the fission reaction takes place.

To quantify the effects of GTRF, before this thesis, a numerical FSI workflow called NRG-FSIFOAM was developed by Nuclear Research Group (NRG). The fluid modelling consisted of a synthetic turbulence model (AniPFM) developed in OpenFOAM. The structural solution was obtained using a 3D Finite Element Method (FEM) solver implemented in deal.II, while the mapping and coupling between the two solvers was handled by preCICE.

In this context, the objective of the thesis is to propose a simplified and cost-effective structural solver, without compromising the accuracy of the methodology. To this end, an eigenmode-based Reduced-Order Model (ROM) of a 1D beam-element FEM formulation is proposed. To further improve the computational costs, by taking into account the 1D FEM formulation of the structural solver, novel mapping routines between the fluid and the structural grid are also proposed. To ensure a strong coupling between the two domains, an Aitken subiteration algorithm is used. What’s more, to simplify the architecture of the methodology, the new features are directly implemented in OpenFOAM. Thus, the newly proposed workflow is fully contained within OpenFOAM, eliminating the need to additionally use deal.II and preCICE. This newly proposed FSI methodology is called NRG-beamFoam.

The thesis first deals with the individual verification of the structural solver, the mapping routines, and the Aitken subiteration scheme. Subsequently, all of the elements are combined within FSI simulations for the same benchmark case that was used for the validation of NRG-FSIFOAM. It is found that for simulations where an Unsteady Reynolds-Averaged Navier Stokes (URANS) fluid model is used, NRG-beamFoam reduces the computational cost per FSI subiteration by 48% compared to its NRG-FSIFOAM predecessor, while obtaining a 0.5% relative difference in the frequency and the damping ratio characterizing the structural dynamic response. What’s more, using the ROM, the instantaneous deformations of a single rod to turbulent excitation by axial water flow appear to be accurately computed using a total number of degrees of freedom that is reduced by a factor of approximately 103 compared to the FEM solver of NRG-FSIFOAM. Further research is recommended to improve the stability of NRG-beamFoam when coupled with the AniPFM for simulation times in the order of seconds. Furthermore, future studies ought to explore the causes for the small amplitude differences observed between the outputs of the ROM and the FEM solver for FSI simulations.